Using Machine Learning to Predict Acute Kidney Injury After Aortic Arch Surgery.

Area of Science:

• Cardiovascular Surgery
• Nephrology
• Artificial Intelligence in Medicine

Background:

• Aortic arch surgery carries a significant risk of postoperative acute kidney injury (AKI).
• Accurate prediction of AKI is crucial for timely intervention and improved patient outcomes.
• Traditional logistic regression models have limitations in predicting complex outcomes like AKI.

Purpose of the Study:

• To compare the predictive performance of machine learning models against traditional logistic regression for AKI after aortic arch surgery.
• To identify the most effective model for predicting kidney outcomes in this patient population.

Main Methods:

• Retrospective review of 897 patients undergoing aortic arch surgery.
• Comparison of three machine learning methods (gradient boosting, support vector machine, random forest) with logistic regression.
• Analysis of perioperative characteristics and assessment of model performance using the area under the receiver operating characteristic curve (AUC).

Main Results:

• 652 patients (72.6%) developed AKI within 7 days post-surgery; 283 (31.5%) had stage 2 or 3 AKI.
• Gradient boosting demonstrated the highest discriminative ability for AKI prediction (AUC 0.8 for binary, 0.71 for multiclass classification).
• Machine learning models, particularly gradient boosting, showed superior performance compared to logistic regression.

Conclusions:

• Machine learning models offer significantly better prediction of AKI following aortic arch surgery than logistic regression.
• Gradient boosting is a highly effective tool for identifying patients at risk of AKI.
• These findings support the integration of machine learning into clinical practice for enhanced perioperative kidney outcome prediction.